Anum's Journey in NLP and Domain-Specific Summarization
Anum shares insight on the progress of her PhD project on Summarization of domain-specific documents.
Hi Anum, please tell us something about yourself.
I'm Anum, a Ph.D. student at the Technical University of Munich. Although my academic journey began in engineering, I discovered my true passion for AI, ultimately becoming an avid enthusiast of Natural Language Processing. I grew up in Karachi, Pakistan’s City of Lights, right next to the Arabian Sea. I believe in being kind, towards others and most importantly towards yourself.
What is your project about and with whom are you collaborating on it?
My project is about summarization of domain-specific documents and I am collaborating with Digital Science. Through my project, I want to investigate finetuning and knowledge injection approaches for Large Language Models.
What passions drive you, aside from your job and the Software Campus?
Beyond my academic pursuits, I'm also a passionate outdoors enthusiast. Whether it's running, biking, hiking, or even a leisurely stroll, I'm always eager to be outside. I appreciate the fresh air because, while it may not solve problems directly, it provides the mental space needed for clear thinking. When I'm not enjoying the outdoors, you'll often find me experimenting and honing my culinary skills in the kitchen, rice & spice are my specialty.
What is it that makes this project / this topic exciting for you personally?
It's a well-acknowledged fact that larger models have performance. Nevertheless, training or employing such models demands substantial resources. Far from being sustainable, right? As our models continue to grow in size, it's crucial to remember that Earth's resources remain finite. I'm particularly enthusiastic about my project with Software Campus because it focuses on enhancing the summarization of domain-specific documents while being mindful about resource utilization.
What is the benefit of a collaboration in this project between academia and industry?
Cutting-edge Large Language Models (LLMs) excel when handling standard data, yet encounter challenges when applied to domain-specific information. Many companies aim to employ LLMs for their internal use cases using their domain-specific datasets. My project seeks to bridge this gap by emphasizing efficient and resource-light techniques for integrating domain-specific data into the model.
How is the Software Campus Program contributing to your personal development?
Thanks to Software Campus, I've had the opportunity to connect with and build lasting friendships with fellow Ph.D. students from across Germany who share my interests. Moreover, in contrast to exclusively concentrating on theoretical research, Software Campus equips us for life beyond our Ph.D. studies by offering diverse training opportunities and fostering collaboration with partner companies.